subagent-driven-development

Dispatch fresh subagents per task with post-task code reviews.

Updated Aug 23, 2026
One-click install
npx skills add https://github.com/HTRamsey/claude-config --skill subagent-driven-development-htramsey
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: subagent-driven-development
Source: https://github.com/HTRamsey/claude-config/tree/main/skills/subagent-driven-development
Command: npx skills add https://github.com/HTRamsey/claude-config --skill subagent-driven-development-htramsey

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Speeds up development by running each task in a fresh subagent and reviewing between tasks to prevent context pollution.

Core Features & Use Cases

  • Sequential vs parallel execution with per-task subagents
  • Code-review after each subagent's work
  • Final review of entire plan

Quick Start

Plan tasks, spawn subagents per task, review results after each, and iterate until complete.

Frequently Asked Questions about subagent-driven-development

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prevent context pollution when running multiple development tasks in parallel?

Context pollution occurs when a single agent accumulates state across unrelated tasks, degrading reasoning quality. Subagent-driven development spawns fresh subagents per task, isolating context and maintaining clarity. Each subagent works independently, then results are reviewed before moving to the next task, ensuring clean reasoning throughout execution.

Can I execute independent tasks in parallel while maintaining code quality?

Yes. This Skill supports wave-based parallel execution where independent tasks run simultaneously in separate subagents, with mandatory post-task code reviews. Configure sequential or parallel modes depending on your workflow, and each subagent's output undergoes review before integration, maintaining quality gates across all execution paths.

What's the best way to structure a multi-task development plan with mandatory reviews?

Break your plan into independent tasks without shared state or dependencies (3+ tasks ideal). Dispatch each to a fresh subagent, collect output, trigger a code review, then proceed to the next task or wave. This Skill automates the dispatch, review triggering, and reporting, with configurable sequential or parallel execution modes and final plan-level review.

How do I handle code review between subtasks without losing momentum?

Subagent-driven development embeds post-task reviews into the execution flow. After each subagent completes work, a review phase runs automatically before the next task begins. This staged approach prevents accumulation of unreviewed code and keeps iteration tight, especially effective when tasks are independent and can be executed sequentially or in parallel waves.

When should I use independent subagents instead of a single agent for development?

Use independent subagents when you have 3+ independent issues or tasks without shared state, context dependency, or coupling. Single-agent approaches accumulate context across disparate work, reducing reasoning quality. Fresh subagents per task excel in execution plans where isolation prevents pollution, task-level reporting is required, and mandatory reviews are critical for quality.

Does this approach work with existing version control and testing workflows?

Yes. This Skill integrates commits and tests into per-task execution, supporting existing CI/CD and version control patterns. Each subagent's work is tested, committed, and reviewed before the next task, maintaining compatibility with standard development practices while enforcing isolation and quality gates.